Hi everyone.

I now have release 1.0 Mk1 and Mk2 volume adjusting kernels available for download. The source patch is also there for people who like to build their own kernels.

My units appear to function fine with these kernels, but I haven't tried everything on the Mk2 yet.

They're at the usual place: http://home.eric.net.au/~rjlov/compress.html.

Richard.

Thanks Richard!

Everyone here appreciates the work you put in on these kernels.

Alex Lear

Thanks! This rocks, just like the previous ones did. It's nice to have 1.0 with the volume adjustment back in.

___________
Tony Fabris

Things haven't changed on the MK2, it's still "10000" as far as I know.

However, do you know whether your copy of Download.exe is the latest version that's Mark2-compatible?

Let me know if that's the source of the problem and I'll emphasize it in the FAQ.

___________
Tony Fabris

Thanks, started to like it... Now its back.

TommyE

Yes, this sounds like the problem: code 1a2 is "page locked" - the flash in the mk2 needs unlocking before any writing is allowed, and the mk1 downloader didn't do this.

Hugo

Hmm.. any chance of having a correct downloader anytime soon? (actually, I can't even find the original download.exe that used to bo on bobo's site - his site's just got a picture up on it now..)

-mark

...proud to have owned one of the first Mark I units

Cool. Yeah, Debauch probably got the .exe from Bobo's site, and I think Bobo has the old one up.

Does anyone else have a more recent compilation of Download.exe that I can link from the FAQ?

___________
Tony Fabris

Hmm.. any chance of having a correct downloader anytime soon?

You could compile it yourself, the source code is linked in the FAQ...

___________
Tony Fabris

Attached is a copy of download.exe that I just compiled myself. Let me know whether it works with the Mark 2, and I will figure out a way to put it in the FAQ.

___________
Tony Fabris

Yep, there's a chance that we may.

Rob

Cool. Di also confirmed this for me, too. I'm working on fixing it so that it'll work on other COM ports as well. Once I find out if that works or not, then I'll post it and link it to the FAQ.

___________
Tony Fabris

Okay, here's a zip file that's got four different versions of download.exe- one for each COM port. I know I could have made the COM port a command-line parameter but I'm lazy and my wife wants me to come get her for lunch, so I'm in a hurry.

I'm going to link this file in the FAQ, too.

___________
Tony Fabris

i know that this is a really stupid question, but any way to tell if this is working? i'm not sure that i can tell the difference between the original 1.0 and the volume adjustement 1.0. the download seemed to work properly. maybe 'cause i'm listening to it at home, and haven't listened to enough songs yet.

jeremy

12 gig, green...

I know that this is a really stupid question, but any way to tell if this is working? i'm not sure that i can tell the difference between the original 1.0 and the volume adjustement 1.0.

Here was my reply to that same question the last time it was asked.

___________
Tony Fabris

Once that volume-adjusting kernal is installed, is there an option to disable/enable it?

I can just see myself in an IASCA competition in the sound linearity test, where the same track is recorded three times, once at -12db, once at 0db, and once at +12db, the object being to see if your spectral balance remains unchanged on playback. Imagine the judges faces when all three tracks play back at the same volume!

tanstaafl.

"There Ain't No Such Thing As A Free Lunch"

Once that volume-adjusting kernal is installed, is there an option to disable/enable it?

No, not on the fly. It's a hack, so it's not integrated with the user interface at all. In the future, if Empeg includes it officially, there will be menu options for it.

Hey, Richard, is there any way to alter the kernel to offer a choice menu at bootup time before the player software is loaded? You could choose to have the feature on or off that way. But once it's into the Empeg player software, there's no way to get out and do anything without a serial connection to a PC.

___________
Tony Fabris

OK, this is a little off topic, and may be a stupid question, but it's been floating around inside my skull -

How does the volume adjustment kernel handle crescendos and decrescendos if there isn't any background music for reference? For example, Metallica's Unforgiven - wouldn't the crescendo at the beginning come out as having a constant volume? Just curious.

Bill B.
Mk.2 SN 080000183 - 18 GB / Green

How does the volume adjustment kernel handle crescendos and decrescendos if there isn't any background music for reference? For example, Metallica's Unforgiven - wouldn't the crescendo at the beginning come out as having a constant volume?

Just like any compressor, it ends up behaving differently depending on the character of the wave data it's being fed. Some fade-ins and fade-outs sound very natural, others have a very flattened volume curve so that they do sound almost constant in volume. It depends on where the peaks lie in the track- I don't own that song so I don't know.

But that's the whole point of a dynamic compressor: to compress the dynamics. In our case, to compress it enough so that the quiet parts come up above the car's noise floor, or so that the quieter songs sound closer to the volume of the loud ones. Yes, in some cases this would tend to destroy crescendos, but that's a choice you have to make: Do you want the first half of the crescendo buried beneath the car's noise floor, or do you want to hear the whole thing? To hear the whole thing, you either have to play it really loud or you have to compress it.

Right now, the parameters of this program are hard-coded. When it gets integrated with the player software, the parameters will be adjustable, and you'll be able to tweak it so that it behaves the way you want it to.

___________
Tony Fabris

I see your point. Thanks for clearing that up for me (I don't really know that much about dynamic compression).

I've only had my player for a couple of weeks now, and most of that time has been spent getting used to it, and setting everything up the way I want it. Looks like I'll have to give the kernel a try pretty soon.

Bill B.
Mk.2 SN 080000183 - 18 GB / Green

I too wondered if it worked the first time I installed it.
Tony assured me that it most likely worked even if I couldn't hear it.
What I did was to take the Empeg for a long drive.
I didn't find myself adjusting the volume so much on that trip.

Play a while with it, and it will become clear.

TommyE

It's definately subtle.. I can usually tell when it raises the volume of something, but I'm not quite so sure if it works for the main thing I need it for - changing the volume BETWEEN songs... example:
really soft song, followed by a really loud song..

The loud song still requires me to turn the volume down manually:(
-mark


...proud to have owned one of the first Mark I units

This may be an absolutely moronic question, but I didn't see anything in the FAQ, on Richard's website, or in this thread. I've normalized my entire CD collection using AudioCatalyst and used the highest VBR setting. Is there any benefit to running this volume adjusting kernal for me?

(Is this one for the FAQ?)

George

I've normalized my entire CD collection using AudioCatalyst and used the highest VBR setting. Is there any benefit to running this volume adjusting kernal for me?

Your question seems to come from two places:

1) Misunderstanding of the difference between "Normalization" and "Dynamic Compression".

2) Misunderstanding of the dual-usage of the word "Compression".

Let's take the second one first:

The word "Compression" is used to describe two different things: The data compression process by which a WAV file is made smaller and turned into an MP3, and the completely unrelated dynamic compression that's used in the original audio production of the albums you listen to and the television shows you watch.

I assume you already understand what data compression means, let's move on to dynamic compression. From here on out, I'm going to refer strictly to dynamic compression.

What it it? It's like a "contrast" control for the sound. A compressor changes the relationships between the loud sounds and the quiet sounds. It can make the loud sounds quieter and the quiet sounds louder, giving the overall volume a smoother flow.

This differs from normalization in that normalization simply adjusts the entire volume of the track upwards until the loudest peak in the track hits 100 percent. Most of the CDs you buy are already normalized, so hitting the "normalize" button in AudioCatalyst rarely makes major changes.

Almost all of the recorded sound you hear, whether it be on TV, movies, or record albums, is already compressed to some degree. And usually multiple times. Let's take radio stations as an extreme example: When you listen to a song on the radio, the sound has already gone through several compressors at different stages. First, the individual instruments and voices were compressed to different degrees when they were recorded, each instrument with different settings. Then, during the mixdown and engineering of the album, some of the instruments were submixed and bounced down to combined tracks (such as the drums being combined into a single stereo track), and were probably run through another compressor in the process. The final mix was probably compressed a bit before it was sent to the mastering facility, and the mastering facility probably applied even more compression to the album. Then when the radio station plays the song, they heavily compress all of the music they play so that all of the songs appear to be at the same volume. This last one explains why songs sound different on the radio than they do when you buy the CD.

Another example is television commercials. They always seem to be louder than the TV shows because they are overcompressed. Both the TV shows and the commercials have the same volume peaks, it's just that the commercials are compressed so they hit that peak more often (almost constantly, in fact). This is completely deliberate on the part of the makers of the TV commercials. The only way around it (and some TV sets will do this for you) is to overcompress the TV shows themselves so that they're now as loud as the commercials.

A great discussion of compression and how it relates to the relative volume of CD songs is right here. Read it, it's terribly informative and it's written by someone whose job depends upon being very good with compressors.

Now, how does this relate to Rjlov's volume adjusting kernel?

Well, Richard's basically written a digital compressor with a very long attack time (although there are differences which I'll get into in a moment). It runs in real time on the songs playing in the Empeg. Its job is to increase the volume of the quiet parts so that they come out closer to the loud parts. The benefit of running this program is that it'll allow you to hear the quietest parts that were originally buried in the noise floor of the car.

It behaves slightly differently than a compressor, in that it doesn't use a fixed "slope" between the input voltage and the output voltage. Instead, it slowly increases the absolute volume towards a blind target until it clips, then instantly backs down if it detects a clip coming up.

This makes it sound a bit different from a straight compressor, but the end result is similar. I like it a little better. If you want to play with a real compressor, load up a wave file into Cool Edit and invoke their compressor filter. You'll be able to see how it works in their user interface and hear the differences.

One thing that Richard still needs to work on is its ability to handle mid-level volume passages. A passage with a medium volume isn't amplified as agressively as a very quiet passage. So the old problem of all your CDs sounding like they're at different volumes is mostly still there. Since I don't currently have the ability to alter the kernel's parameters, I don't have any way of experimenting to see if that's adjustable. If this software ever gets integrated with the Empeg's main user interface, then we'll be able to play with the settings and get different results. You have to be careful of this sort of thing, though, because an overcompressed album sounds bad.

Does this answer your question?

___________
Tony Fabris

Would it be fair to suggest that if this is included in the official empeg software, that it should only be enabled when the empeg is in the car, and disabled if in a home environment?

At least it should be configurable to be enabled/disabled separately for home/car use. If I understand it correctly (from posts here and from the link Tony posted to the explanation of dynamic compression) then having the volume adjustment permanently enabled in both home and car environments is not necessarily ideal.

Geoff
---- -------
Got one of the first Mark 2 empegs...

Of course. The reason we'd like to see it integrated with the player software is so that we could turn it on and off and adjust its parameters.

You're right in that its main usage would be in the car when you're driving at highway speeds. But as stated in that article about compression (the one I linked in my compression explanation), there are many in-home situations where you might want it as well. That's why the author of that article suggested that consumer products should have a "party" button on them, because that's one of the places you'd want to use it.

Another thing that was discussed in the Reviews forum was the idea that it could possibly use the mic input to detect ambient noise, and increase the agressiveness of the compression as the road noise increased. That would be awesome.

Still, your suggestion that it have separate home/car personalities is a good one. Empeg guys: As you're designing the plug-in specification for this kind of thing, make sure that home/car personality detection is part of the API, K?

___________
Tony Fabris

OK. There's a few questions that have been asked that I shall try and address.

- Does the volume adjustment destroy cresc/dims?

No, but it will usually (almost always) make them less pronounced. It attempts to preserve volume relationships. In a steady state, if a > b, then voladj(a) > voladj(b). So, you should still get all your dynamics, just reduced in contrast. But, it is a gradual thing, so if there is a cresc faster than about 3db/second, you will get a cresc then dim effect. In practice, these are fairly rare.

- How come I still have to twiddle my volume knob?

Not all CDs are normalised. But in my experience very few of them have the largest sample less than 6db below the maximum possible sample. With the current volume adjustment this pretty much translates to a 2db drop from the loudest track you can play. That's not much, and I notice that many of my tracks no longer need manual adjustment, but see below.

But, there's this headroom multiplier I put in there (it's 0.75 at the moment). It's supposed to make clipping happen less often, but one of the side-effects is that only the stuff that is already really loud can be loud after the volume adjustment. It's a little bit complicated, but I'll try and explain.

outval = multiplier( inval ) * inval
multiplier( inval ) = max( 1, headroom * desired_output( inval ) / inval )

The max(1,...) is there because there is no point in having the multiplier less than 1. In that case we're just throwing away information. But the max() also means that the linearity (or logularity??) is broken when the desired_output(inval) / inval > 1 / headroom.

How can we fix it? Well, we can try using headroom = 1. Then this effect will disappear, and you will have much less need to adjust the volume for tracks from different CDs, but you are more likely to get objectionable, rapid volume changes as the routine avoids clipping. How much more likely? I don't know. Try it and see! You can use the command line version to experiment.

I think the next version I make I'll try having headroom = 1.

The other solution is to normalise all your CDs before encoding them, which I recommend (but I don't do, yet).

- How do I know if it's working?

Most of the time I can't tell either. But, if you have tunes with a really soft passage in the middle (that you usually don't hear) you might notice that it now is audible. Another time is when there is interesting stuff happening at the end of a fadeout, that is much louder than normal. If you have recordings with long periods of silence, but a high level of background noise (enough to trip the silence thresholds) (usually older recordings) then you will notice the hiss being brought up in the silence, and it dropping away again when the music starts.

Basically, if you can't tell it's there, that's what's intended.

- Can I turn it on and off, or adjust the parameters?

No. I would like this as much as you would and when we get an API for this kind of thing (it's in the works, I'm told) we'll have some user interface for it in the player. It probably wouldn't be that hard to activate / deactivate it based on home / car in the current form, but I like it at home as well! If the music is just background, then the volume adjustment is good. If you're actually sitting there concentrating on it, then you don't want it.

It would be nice to switch it on and off so you can see how much benefit (if any) it is providing.

Ideally we could get the frobbing knob on the Mk2 to adjust the minimum desired volume, then you just dial up no distortion for Doug, or no contrast for when the ambient noise floor is at the threshold of pain. :)

Richard.

If you have tunes with a really soft passage in the middle (...) you might notice that it now is audible
you can't tell it's there, that's what's intended.

Yeah . . . . Great stuff
You're right: one can hardly notice that it's there, but soft passages are audible now ! Really clever.
Thanks for sharing it, Richard. This must be the type of player extensions that Hugo at al want to trigger.
Now let's hope the empeg guys will find the time so it can be neatly integrated in the player. Both the player and your extension deserve it.

Henno
mk2 6 nr 6

Another thing that was discussed in the Reviews forum was the idea that it could possibly use the mic input to detect ambient noise, and increase the agressiveness of the compression as the road noise increased. That would be awesome.

<delurk>
Even better - generate an inverse signal for noise cancellation. Would that work? Now that would be incredible.
<relurk>

Borislav

This is something that has crossed my mind from time to time too.
Don't know if something like this WOULD work?
Anyone who have any thoughts???


TommyE

Noise cancellation is a huge problem. Many auto manufacturers have attempted something like this. It is a very sensative process. The biggest problem is you have to monitor the placement of your ears (not really as necesary in a car) to adjust the phases for every frequency across the spectrum. Another problem is a cancelled noise for the person in the front seat may be doubled in amplitude for a person in another seat. These problems have no effect for noise cancelling headphones, this is why they work as well as they do.

Alex Lear

All of the audio settings have two personalities (home and car) and I don't see why dynamic compression would be any different.

Rob

Noise cancellation is an established technology, but it tends only to be really effective in very specific or controlled environments. You can cancel a lot of prop noise in a small plane because you know the cabin dynamics, but cancelling generic noise in a generic car cabin would be an entirely different kettle of fish.

I believe that some auto manufacturers are researching this (maybe there are even early systems on the market, I haven't checked recently) but they have an advantage because they're designing for a specific model and audio system.

Rob

All of the audio settings have two personalities (home and car)

Not entirely true, Rob:
There's only a single loudness setting (home and car), which IMHO doesn't make sense.
Bug or oversight?


Henno
mk2 6 nr 6

Oversight I guess.. report it! :-)

(I know you just did, but if you email it I won't have to try to remember it in the morning!)

Rob

I envisioned having the volume pop up screen with both a minimum and maximum bar (basically the same as now, with another bar next to it), and then being able to adjust those. Anyone think this would confuse people?

Er, I'm still probably not explaining this correctly, it's like this http://rmitz.org/example.gif. The text would presumably also have to be changed.

Fly me to the moon...

Interesting idea, but there's more parameters to a compressor than just the minimum threshold.

And technically, the compressor has nothing to do with the volume. It just compresses the source wave data without altering the volume setting.

___________
Tony Fabris

I know that this isn't quite how the current version of the code works, but as rjlov said, this is the model he wants to move toward. Unless I misunderstood him, of course.

Fly me to the moon...

Lexus has done work with noise cancelation a while ago. While they can't cancel *any* noise in the cabin, they can cancel the engine noise actively and passively. They use noise absorbing materials to reduce noise down and then after that they can cancel most of the remaining engine noise actively.

Calvin

Does this answer your question?

Yeah, and then some! Thanks!

I do have a question about normalization in Audio Catalyst. In its' normalize option, it says "Normalize to [a percentage]" and then says "but only if the track is lower than [a percentage] or higher than [a percentage]"

Does that "lower than" have anything to do with bringing the noise floor up? These are the setting that confused me as to how voladj would help. It SOUNDS (from reading the options) like it's bringing the noise floor up and the peak back down. Is this not the case? If this *is* the case, isn't this what you just described, except that voladj does it in real-time? Perhaps there's more going on than only this in voladj, but that's the BASIC (not to diminish the work on it!) piece, isn't it?

George

normalization in Audio Catalyst

Hi George,
I too was confused by the AC options when I started to rip for my empeg. Now I find the AC options for 'Variable Bit Rate' and 'Normalisation' very useful. This is the way I believe it works:

Normalisation will adjust (bring up or down) the output level of the entire track by the same amount. To do this, AC will listen to the entire track and find the peak recording level and, if this falls within the boundaries that make normalisation worthwhile, adjust the volume of the entire recording equally. I like to think about this as the binary representation of floating point numbers: I imagine that normalisation leaves the fraction alone (content remains unchanged) but adjusts the exponent (the multiplier i.e. volume) of the frames such that the peak equals the percentage of max output that you set.

I now have this normalisation option on all the time when I rip for the empeg player and have noticed that most new CD remain untouched: they don't need adjustment: almost all are mastered in a such way that max output is at 98% or 99%. Older stuff, however, is sometimes put on CD at much lower levels, and therefore sounds less loud too. To listen to these CD's you'd need to adjust the volume of your head unit / pre-amp. Normalisation will allow to bring such older recordings up to the same volume level as the new stuff, so that you won't have to adjust the volume of your empeg player (as much) between tracks. In theory you ask AC to normalise to any level, so you can also normalise recordings down to say 70% or 50%, but I see no benefit in doing this. During play-back you'd need to amplify these more / introduce more distortion.

My understanding of Richard's 'Volume Adjust' is that it looks ahead to see if the output level of what's going to be played next is significantly different from what's currently playing and, if so, smoothes out the differences. As a result, low volume passages are turned up a bit and when if finds sudden increases in output level, it quickly turns down the volume to avoid clipping. (I don't know if Richard manipulates the empeg volume, of the output level data in the recording. I would think the latter.)

In this way Richards utility complements normalisation: AC's normalisation will smooth out differences between recordings, but leaves the soft/loud differences in tact. Richard's 'Volume Adjust' will smooth out the differences, such that the low passages are not lost in the back ground noise.

Henno
mk2 6 nr 6

Does that "lower than" have anything to do with bringing the noise floor up?

I'm afraid not. The text in the dialog box is misleading.

Currently it says: "But only if the track is lower than (x) or higher than (x)", When it should actually say: "But only if the TRACK'S PEAK is lower than (x) or higher than (x)".

All it does is look at the single highest peak in the track, then add a multiplier to the whole track based on those criteria. Personally, I normalize everything unconditionally to 100 percent. Although it helps some older albums that weren't fully normalized, for the most part it does nothing because most albums are already normalized. I'd say that out of my entire CD collection, there were maybe two or three CDs that benefited from this.

Normalization doesn't usually fix the perceived-volume differences between tracks because most of those differences are due to differences in the amounts of compression in the mastering stage.

___________
Tony Fabris

Okay, I think I finally understand this. Richard's program adjusts it dynamically, meaning it raises the lows ONLY and raises the the highs ONLY, when necessary. Audiocatalyst does both of those, regardless and not in proportion to the part that's playing, it's adjusted the same amount throughout the entire track. If this were drawn on a graph, you'd have the track shown as a straight bar across the middle, representing the original sound. Audiocatalyst would have two identical lines running the length of the track, one across the top and one across the bottom, with the peaks and pits at the same points on both the top and bottom. Voladj would have two independent lines across the top and bottom, with the peaks and pits at different points.

I think that makes sense. If not, it does to me!

George

Just a "Me too". I have downloaded the V1 vol adjust kernel to my Mk I and it is great. thanks for that.

PeterH
" 'tis better to be silent and thought a fool than to speak and remove all doubt"